Download IBIS BEHAVIORAL MODELS TECHNICAL NOTE

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Document related concepts

Immunity-aware programming wikipedia , lookup

Opto-isolator wikipedia , lookup

Transcript
TN-00-07
IBIS BEHAVIORAL MODELS
TECHNICAL
NOTE
IBIS BEHAVIORAL
MODELS
This article was originally published in 1996.
STRUCTURE OF AN IBIS MODEL
The current version of IBIS is composed of five basic
elements as shown in Figure 1 (Ref. 1) below. Element
1 is the pull-down structure, element 2 is the pull-up
structure, element 3 contains the protection devices,
element 4 is the intrinsic skew time for a rising or falling
signal, and element 5 is the package parasitics. Within
an IBIS file, each pin can have a different model to
reflect differences in input or output structure.
INTRODUCTION
A new standard for formatting analog models of
digital integrated circuit I/O characteristics is beginning to emerge and promises to provide fast, accurate
simulations for transmission line and signal integrity
analysis. This standard is rapidly being adopted by
semiconductor manufacturers and electronic design
automation vendors to assist system designers in efforts to increase bus speeds with greater confidence.
The I/O-buffer information specification (IBIS) was
developed at Intel as a way of supporting designs for the
Pentium® system. The IBIS specification provides a
standard parsed file format of current/voltage (I/V)
characteristics, device package parasitics and ESD protection device characteristics for several types of I/O
structures. Some of these structures include ECL, open
source, open drain and tristate.
3
2
4
Rise
Fall
5
L_pkg R_pkg
C_comp
1
C_pkg
SPICE VS. IBIS
IBIS models obtained from manufacturers have definite advantages over SPICE models. Most IBIS behavioral data is taken from actual devices. Thus, if the IBIS
data is properly taken, the resulting model data has a
direct correlation to the actual device. Another advantage is that IBIS models tend to simulate much faster
than SPICE models. Some sources quote as much as a
25-fold increase in the performance of IBIS over SPICE.
Why would you need IBIS-formatted models if SPICE
models are available? The answer to this question is
quite simple. Semiconductor manufacturers are reluctant to provide SPICE models of I/O structures to the
public because the models often communicate confidential process and circuit design information or the
manufacturer only has models for in-house siliconsimulation tools. When SPICE models are made available to the public, the models are often “sanitized” so
that any proprietary information is masked. This can
result in a model that does not accurately convey the
analog behavior of the device. Whether or not the
SPICE model is accurate depends on the nature of the
proprietary information removed from the model and
the extent to which the model made available to the
public is maintained to reflect the latest process and
design of the device.
TN-00-07
TN0007.p65 – Rev. 9/99
Figure 1
ELEMENTS OF AN IBIS MODEL
The following are examples of what the elements of
an IBIS model look like when formatted for the IBIS file
parser. Element 1 is the pull-down information that
describes the I/V characteristics during pull-down. You
can see that there is data for minimum and maximum
current for given voltages. (These are temperature- and
process-dependent specifications.)
[Pull-down]
|Voltage
I(typ)
|
-5
-177.7m
-4
-149.5m
-3
-117.5m
-2
-81.4m
data omitted . . . . .
10
139.4m
I(min)
I(max)
-165.8m
-138.8m
-108.5m
-74.5m
-189.6m
-160.1m
-126.5m
-88.3m
128.6m
150.2m
You may notice that the data is taken for -VDD to
2VDD. The main reason the voltage sweep is done over
such a wide range is that this allows a behavioral model
1
Micron Technology, Inc., reserves the right to change products or specifications without notice.
©1999, Micron Technology, Inc.
TN-00-07
IBIS BEHAVIORAL MODELS
for signal reflections caused by improper termination.
Sweeping over this voltage range also allows modeling
of the device for overshoot and undershoot situations
when the protection diodes are forward biased. The
current from the protection diodes is not included in
this data. When the data is captured, the current from
the protection diode is subtracted from the overall
current before the current data is entered in the table.
Diode current is later added by the simulator from
model data when appropriate.
Element 2 contains data for the pull-up state of the
buffer when the output drives high. This data is entered
into the table using the formula VTABLE = VDD - VOUTPUT
since the current through the pull-up device is dependent on the voltage between the output and VDD.
[Pull-up]
|
|Voltage
I(typ)
|
-5
-223.0m
-4
-197.4m
-3
-161.8m
-2
-125.3m
data omitted. . . . .
10
0.0
I(min)
I(max)
-204.0m
-190.7m
-156.6m
-118.6m
-242.0m
-204.0m
-167.0m
-132.0m
0.0
0.0
10
0.1u
|
[POWER_clamp]
|
|Voltage
I(typ)
|
-5
18.4m
-4
11.6m
-3
3.5m
-2
0.8m
data omitted. . . . .
10
0.1u
TN-00-07
TN0007.p65 – Rev. 9/99
I(min)
NA
NA
NA
NA
NA
I(min)
I(max)
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
The power and ground clamp diodes are modeled in
parallel with the driver information in the first two
elements to ensure that the diode characteristics are
present even when the output buffer is in tristate. Data
for the diodes is taken with the output tristated (if
possible) and is done using the same methods described
above for the pull-up and pull-down structures.
Element 4 contains the ramp time for the pull-up
and pull-down devices. This information is indicated
by the keyword [Ramp] and is provided to ensure
proper AC operation of the model.
[Ramp]
|
dV/dt_f
dV/dt_r
The minimum and maximum values are determined
by the minimum and maximum operating temperatures, supply voltages and process variations. With the
minimum and maximum values for data points, a
worst-case and best-case model can be assembled. By
combining the highest current values with the fastest
ramp time and minimum package characteristics, a fast
model can be derived. A slow model can be derived by
combining the lowest current with the slowest ramp
time and maximum package characteristics.
Notice the keyword surrounded by square brackets
before each element of the model. This keyword indicates the nature of the data that follows so that the
simulator can either use the data from the parsed ASCII
file directly or translate the data into a format used by
the simulation software.
Element 3 contains data describing the ground and
power clamp diodes.
[GND_clamp]
|
|Voltage
I(typ)
|
-5
-188.3m
-4
-144.9m
-3
-109.3m
-2
-58.9m
data omitted. . . . .
NA
typ
min
1.98/58p 1.87/60.5p
1.55/158p 1.46/160p
max
2.09/56.9p
1.64/177p
The min and max columns represent the minimum and
maximum slew rates for the buffers. These values are
very small since they represent the intrinsic values of
the transistors with all package parasitics and external
loads removed.
Element 5 adds the component and package
parasitics. C_comp is the capacitance of the die itself,
excluding the package capacitance. The package characteristic resistance, inductance and capacitance are
added by R_pkg, L_pkg, and C_pkg, respectively.
[Package]
|
|
typ
min
R_pkg
52mohm 51mohm
L_pkg
6.425nH
3.33nH
C_pkg
0.875pF
0.4pF
•
data omitted. . . . .
•
|
typ
min
C_comp
2.33p
2.17p
I(max)
NA
NA
NA
NA
max
53mohm
9.52nH
1.35pF
max
3.14p
Many of the capabilities of the IBIS model to simulate “real world” situations are actually dependent on
the capabilities of the simulation software. The latest
2
Micron Technology, Inc., reserves the right to change products or specifications without notice.
©1999, Micron Technology, Inc.
TN-00-07
IBIS BEHAVIORAL MODELS
version of the IBIS format (Version 3.0) contains more
keyword areas that provide necessary data to model
such things as ground bounce, simultaneous switching
output events, gradual turn on (slew rate controlled)
outputs, etc. IBIS models cannot be used to measure
propagation delays or provide timing information. The
main purpose of IBIS data is to simulate transmission
lines and analyze signal integrity issues.
Through this process, a majority vote of the forum
ratifies proposed changes.
The latest IBIS version as of this writing is version
3.0. All versions (future and present) are backwards
compatible with previous versions. Version 3.0 has
been ratified as an EIA standard.
The IBIS committee is constantly working to improve the quality and quantity of information provided in IBIS standard files. New versions of the standard file format are constantly under review.
IBIS OPEN FORUM
Analog simulator vendors, computer manufacturers
and IC vendors have formed a group known as the IBIS
Open Forum. This group is an open, voluntary association of companies that have worked to develop the IBIS
standard in its present form. At the time of this writing,
the group has managed to become part of the Electronics Industries Association and is working to get IBIS
accepted as an EIA standard. This group has also assembled documentation (cookbook) outlining suggested
methods for data extraction. Using the methods outlined in the “IBIS Cookbook,” a model with the five
elements shown in Figure 1 can be assembled.
There is a program available from the IBIS forum
called the IBIS Golden Parser that checks an IBIScompatible model for proper syntax. An IBIS model
that follows the proper syntax can be used either
directly by a simulator or translated into a format that
can be used by a simulator. This program is available on
several platforms and is offered at no charge by the IBIS
forum. The source code is available for a nominal fee.
Not only can the data be extracted from empirical
data, but a program developed at North Carolina State
University translates SPICE to IBIS. The program works
with Berkeley SPICE 2 and 3, PSPICE and HSPICE. This
translator produces IBIS-formatted output from actual
simulation results. The program is available in the
public domain for semiconductor manufacturers and
end users to use to generate IBIS models from SPICE
models.
SOFTWARE VENDOR SUPPORT OF IBIS
As of this writing, there are 33 companies and
organizations that have either participated in creating
the IBIS specification or are creating/distributing/using
IBIS models. As more models are made available and the
IBIS format becomes more widely known, the number
of software vendors supporting IBIS will increase.
MICRON SUPPORT OF IBIS
Micron has already produced IBIS models for several
DRAM and SRAM devices; IBIS models for Flash devices
are available upon request. Future support of IBIS will
continue as long as we continue to receive customer
requests for this type of assistance.
MORE INFORMATION
More information about IBIS and the IBIS Open
Forum is available through the Internet. To subscribe to
the IBIS Forum reflector, send your e-mail address to
[email protected]. The domain is 198.31.14.3. For
more information, visit their Web site at www.eia.org/
eia/ibis/ibis.html.
Information is also available through anonymous
ftp at vhdl.org or by calling the VHDL International
BBS (414) 335-0110. You will find documentation, IBIS
Cookbook, recent press releases, Golden Parser, etc.,
residing in the pub/ibis directory.
REFERENCES
FUTURE VERSIONS OF IBIS
Duehren, D., W. Hobbs, A. Muranyi, and R. Rosenbaum. 1995. “I/O-buffer
Modeling Spec Simplifies Simulation for High-speed Systems,“ EDN. March
16, 1995: 65-70.
The process of proposing changes to the IBIS format
is through a buffer issue resolution document (BIRD ).
8000 S. Federal Way, P.O. Box 6, Boise, ID 83707-0006, Tel: 208-368-3900
E-mail: [email protected], Internet: http://www.micron.com, Customer Comment Line: 800-932-4992
Micron is a registered trademark of Micron Technology, Inc.
Pentium is a registered trademark of Intel Corporation.
TN-00-07
TN0007.p65 – Rev. 9/99
3
Micron Technology, Inc., reserves the right to change products or specifications without notice.
©1999, Micron Technology, Inc.